Disjoint reference in real-time comprehension: Computational and cross-linguistic perspectives

对于说母语的成年人来说，语言理解通常似乎毫不费力。但这种印象掩盖了这样一个事实，即实时理解语言需要大量的潜意识脑力劳动。语言理解中的一项困难任务是理解代词。代词是指“他”或“她”等指代不足的词，它们能在上下文中理解它们的意思。为了理解像‘Ellie Like Her’这样的简单句子中的‘Her’指的是谁，听众依赖于一组复杂的语言和非语言线索。语言线索似乎并不是武断的：语言学家已经发现，在语言如何限制代词可能的含义方面，存在着强烈的跨语言倾向，不同语言之间的差异有限。这种跨语言的相似性通常被认为是反映了说所有语言的人所共有的人类认知方面。因此，研究代词理解中的跨语言相似性和变异性为了解人类大脑如何使用语言创造意义提供了一个窗口。更好地了解说不同语言的人是如何理解代词的，反过来可以帮助治疗语言障碍，以及开发自然语言处理系统，使计算机能够更好地接近人类语言的使用。这个项目将用跨语言的心理语言学方法来研究这一问题。研究小组将使用眼球跟踪来衡量听者如何实时理解代词。特别是，他们将调查语言上对代词指称的限制是如何被听者实时使用的：一种语言对代词意义的语法限制是立即影响理解，还是只在以后的阶段限制解释？为了探索这个问题，本项目将调查在关键语法特性不同的一些无关语言中的代词理解，并使用计算实现的认知模型来理解结果。该项目将包括为不同职业阶段的研究人员提供高级STEM培训，从本科生到博士后学者。该项目的跨语言部分将涉及为语言学方面代表性不足的群体成员提供培训和外联机会。研究团队还将开发视频材料，使公众能够了解这项研究的结果。这一奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Language comprehension generally seems effortless for adult speakers in their native language. But this impression masks the fact that a significant amount of subconscious mental work is necessary to understand language in real-time. One difficult task in language comprehension is understanding pronouns, referentially underspecified words like 'he' or 'she' that pick up their meaning in context. In order to understand who 'her' refers to in a simple sentence like 'Ellie likes her,' listeners rely on a complex set of linguistic and extralinguistic cues. The linguistic cues do not seem to be arbitrary: Linguists have identified strong cross-linguistic tendencies in how languages constrain the possible meanings of a pronoun, with a limited amount of variation across languages. This cross-linguistic similarity is generally taken to reflect aspects of human cognition that are shared by speakers of all languages. Thus, studying cross-linguistic similarity and variation in pronoun comprehension offers one window into how the human mind creates meaning using language. A better understanding of how speakers of different languages understand pronouns can in turn help in the treatment of language disorders, as well as in the development of natural language processing systems to allow computers to better approximate human language use.This project will investigate this with a cross-linguistic, psycholinguistic approach. The research team will use eye-tracking to measure how listeners understand pronouns in real-time. In particular, they will investigate how linguistic constraints on pronoun reference are used by listeners in real-time: Does a language’s grammatical constraints on a pronoun's meaning immediately influence comprehension, or does it only restrict interpretation at a later stage? To explore this question, the present project will investigate pronoun comprehension in a number of unrelated languages that vary in key grammatical properties, and use computationally implemented cognitive models to understand the results. This project will involve advanced STEM training for researchers at different career stages, from undergraduates to post-doctoral scholars. The cross-linguistic component of this project will involve training and outreach opportunities for members of underrepresented groups in linguistics. The research team will also develop video materials to make the results of this research accessible to the general public.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.